Transcript Climate Change - Sensible Policy

New Developments in Global
Economic Modelling
Warwick J. McKibbin
Centre for Applied Macroeconomic Analysis ANU
& Lowy Institute for International Policy, Sydney
& The Brookings Institution.
Presentation prepared for the QUT Symposium on Contemporary Challenges in
Economic Modelling and Forecasting, in Brisbane 16 July 2004
1
Overview
• What are multi-country models?
• What are they Useful for?
• Types of Global Economic Models
– Key Features
– Strengths and Weaknesses
• A closer look at a new generation of models (dynamic intertemporal
general equilibrium models)
• Using global models for Policy and scenario analysis
• Research Plans/Challenges
– Model development
– Software development
– Training
• Conclusions
2
What is an Empirical Economic Model?
• A set of equations embodying the history of theoretical
and empirical economic knowledge
• Key bits
– identities
– behavioural equations
– exogenous inputs
3
How to Use Global Economic Models
• Models generate both quantitative estimates and new
insights on complex issues
• Models are useful for
– Forecasting
– Policy evaluation
– Scenario analysis
4
What Features are Important in a Model?
• Does the model explain anything we observe today or in the recent
past? ;
– Model validation is important
• Estimation of parameters
• Replication of history over time and key case studies
• Evaluation of projections or forecasts
•
•
•
•
Is the model continually reviewed by experts who actually use it?;
Is the model published in the refereed academic literature?;
Is there a full listing of all equations available on request?;
Is the model generally open to evaluation by others?;
5
What Features are Important in a Model?
• Is the private market willing to purchase the model for
the value it provides? ;
• Do the model results pass the test of common sense?;
• Are the mechanisms in the model transparent to other
trained economists?
6
Types of Structural Global Models
Structural Global
Models
Input/Output
Models
Computable General Equilibrium
Models (CGE)
Models
Dynamic Intertemporal General Equilibrium
Models
Old Style Macroeconometric
Models
1960s/70s
Modern Macroeconometric Models
1980s/90s
Stochastic Dynamic General Equilibrium
Models
Types of Global Models
• Input Output Models
– Examples:
• United Nations Global models developed by
Wassily Leontieff, Faye Duchin
8
Types of Global Models
• Input Output Models
– Trace the flow of resources between sectors;
– Little role for relative price changes or substitution of
inputs or consumption bundles;
– Tend to be static;
– No allowance for capital accumulation or international
financial flows;
– Ignore the role of money and asset prices;
9
Types of Global Models
• Computable General Equilibrium Models (CGE)
– Examples:
• Domestic
– ORANI model, Monash Model
– Murphy 303
• Multi-Country
– MEGABARE, GIGABARE, GTEM
– GTAP World Trade Model
– Michigan World Trade Model
10
Types of Global Models
• Computable General Equilibrium Models (CGE)
– Derived from microeconomic optimization theory;
– Considerable attention to individual behavior;
– Relatively easy to understand results given theoretical structure;
– Inadequate macroeconomic behavior;
– Tend to be comparative static or recursive dynamic;
11
Types of Global Models
• Computable General Equilibrium Models (CGE)
• Inadequate treatment of intertemporal saving and
investment decisions, capital accumulation,
financial capital flows;
• Ignore the role of money and asset prices;
• Rarely validated with actual experience either
econometrically or through forecasting or shock
replication.
12
Types of Global Models
• Old Style Macro-econometric Models
– Examples
• Domestic
– NIF Treasury Model
– Reserve Bank Models I & II
• International
– Data Resources Inc (DRI)
– Warton
13
Types of Global Models
• Old Style Macro-econometric Models
– Rely on correlations in time series data based on
aggregate economic theory;
– Reasonably good for short term forecasting (tend to
be quarterly)
– Difficult to understand results because of lack of
theoretical structure;
– Unclear long run properties;
14
Types of Global Models
• Modern Macro-econometric Models
– Examples
• Domestic
– Murphy model 2
– Access Economics
– Treasury TRYM model
• International
– IMF Multimod
– Oxford Econometric Forecasting (OEF)
– GEM model of LBS/ NIESR
15
Types of Global Models
• Modern Macro-econometric Models
– More tightly specified theory
– Rational Expectations in some markets
– short run data consistency with long run theoretical
properties
– tend to be quarterly
16
Types of Global Models
• Intertemporal GE Models
– Domestic
• Jorgenson-Wilcoxen US model
– Multi-Country
• The MSG2 Multi-Country Model
» (McKibbin & Sachs)
• The GCUBED Multi-Country Model
» (McKibbin & Wilcoxen)
– GCUBED Environment
– GCUBED (Asia Pacific)
– GCUBED (Agriculture)
– The MSG3 Multi-Country Model
17
Types of Global Models
• Dynamic Intertemporal GE Models
– integrates the key features of the other types of
models
– mix of econometric estimation and calibration
– annual frequency
– problem with large degree of disaggregation because
of complexity of the numerical algorithms needs
18
Types of Global Models
• New Keynesian Models
– Since the publication of the Obstfeld and Rogoff
textbook on International Economics, macro modelers
have also discovered intertemporal models
• Sticky prices
• Mix of optimizing and rule of thumb agents
• Imperfect competition
19
Types of Global Models
• Substantial convergence across the various schools of
structural global modelling
20
Dynamic Intertemporal General Equilibrium Models
21
Overall model development strategy
• Funding is both through research grants and private
consulting
• Hub and spoke approach to coordinating a global
research project (different strategy to Project Link)
– The model is managed/developed in the core
research team in Australia and Syracuse
– Users (researchers/ governments/ financial investors)
in different countries feed back to the core group both
their own developments of the model as well as
funding the core for new developments. All of which
which we are able to incorporate into the model over
time
22
Features of DIGE models
•
•
•
•
•
•
•
Dynamic
Intertemporal
General Equilibrium
Multi-Country
Multi-sectoral
Econometric
Macroeconomic
23
The MSG2 Multi-country model
(1984-1994)
McKibbin and Sachs
24
Development and Subscription Funding
–
–
–
–
–
–
–
–
–
–
–
–
–
McKibbin Software Group Inc
US Congressional Budget Office
The Brookings Institution
US Department of Commerce
US Government
United Nations
World Bank
Australian Treasury
Centre for International Economics
Nomura Research Institute
Daewoo Research Institute (Korea)
Warwick Modeling Bureau
Many Academic Colleagues
25
The MSG2 Model
• Countries
• United States
• Japan
• Germany
• France
• Canada
• United Kingdom
• Italy
• Austria
• Australia
• New Zealand
• China
- Taiwan
- Malaysia
- Indonesia
- Thailand
- India
-Philippines
- Hong Kong
- Singapore
- Korea
26
The MSG2 Model
•
•
•
•
•
1 production sector in each country
macroeconomic focus
International capital and trade flows
Forward looking expectations by some agents
Rigidities in physical capital formation but highly
mobile financial capital
• Unemployment is labour markets due to institutional
factors
27
The G-Cubed Model
(1991- )
McKibbin & Wilcoxen
28
Development and Subscription Funding
– Major Funding
• The Brookings Institution
• United States Environmental Protection Agency
• United States National Science Foundation
• McKibbin Software Group Inc
– Minor Funding through consultancies
• United Nations
• World Bank
• Australian Dept of Environment/AGO
• New Zealand Department of Commerce
• Canadian Dept of Finance
29
The G-Cubed Model
– Countries
• United States
• Japan
• Australia
• New Zealand
• Canada
• Rest of OECD
• Brazil
• Rest of Latin America
• China
• India
• Eastern Europe and Former Soviet Union
• Oil Exporting Developing Countries
• Other non Oil Exporting Developing Countries
30
The G-Cubed Model
– Sectors
–
–
–
–
–
–
–
–
–
–
–
–
Electric Utilities
Gas Utilities
Petroleum Refining
Coal Mining
Crude Oil and Gas Extraction
Other Mining
Agriculture, Fishing and Hunting
Forestry and Wood Products
Durable Manufacturing
Non Durable Manufacturing
Transportation
Services
Capital Producing sector
31
The G-Cubed (Asia Pacific) Model
32
Countries
•
•
•
•
•
•
•
•
•
•
•
•
United States
Japan
Australia
New Zealand
Canada
Korea
Rest of the OECD
Thailand
Indonesia
China
Malaysia
Singapore
Taiwan
Hong Kong
Philippines
India
Oil Exporting Developing Countries
Eastern Europe and the former Soviet Union
Other Developing Countries
33
G-Cubed (Asia Pacific)
– Sectors
• Energy
• Mining
• Agriculture
• Durable Manufacturing
• Non-Durable Manufacturing
• Services
Capital producing sector
34
The G-Cubed (Agriculture) Model
35
G-Cubed (Agriculture)
– Countries
– United States
– Japan
– Australia
– EU12
– Canada
– Mexico
– ROECD
– China & Hong Kong
– ASEAN
– Taiwan
– Korea
– ROW
36
G-Cubed (Agriculture)
– Sectors
• Food grains (rice and wheat)
• Feed grains
• Non-grain crops
• Livestock and its products
• Processed food
• Forest and Fishery
• Mining
• Energy
• Textile and Clothing
• Other non-durable consumer goods
• Durable consumer goods
• Services
Capital Producing sector
37
Agents and Markets
•
•
•
•
AGENTS
Households
Firms
Governments
MARKETS
Goods & Services
Factors of Production
Money
Bond
Equity
Foreign Exchange
38
: The Structure of the G-Cubed Use Table
1
...
12
C
I
G
1
...
A
B
12
R
K
C
Kc
L
A) Interindustry transactions.
B) Industry sales to final demand sectors.
C) Purchases of primary factors by industries.
D) Purchases of primary factors by final demand sectors.
D
X
M
Key Features
• the demand and supply side of the major economies are
explicitly modelled ;
• demand and supply equations are based on a
combination of intertemporal optimizing behavior and
liquidity constrained behavior;
• Explicit treatment of asset markets including money;
• Sticky wages based on labour market institutions imply
unemployment can persist for many years
40
Key Features
• Distinction between stickiness of physical capital within
sectors and countries and the flexibility of financial
capital which immediately flows to where expected
returns are highest
• Extensive econometric estimation of key consumption
and production substitution elasticities
41
Households
– 2 types
• A) maximize an intertemporal utility function
consisting of all goods and services produced
domestically and overseas, subject to an
intertemporal budget constraint that the present
value of consumption is bounded by the
present value of after tax income from all
sources
• B)Base aggregate consumption expenditure
on an optimal rule of thumb with current
consumption of each good allocated so as to
maximize contemporaneous utility
42
Firms
– 2 types
• A) Maximise their share market value (the
present value of the future stream of dividends)
subject to production technology, a cost of
adjustment model of capital and taking prices
as given. They base their calculation on a
summary of the future measured by Tobin’s Q.
• B)Base aggregate investment expenditure on
an optimal rule of thumb with investment equal
to cash flow after paying for variable factors of
production
43
Governments
• Governments provide public goods that enter
into the utility functions on households
(additively separable) and transfer payments;
• They collect a wide variety of taxes on income
of firms households, imports, sales.
• Governments are subject to the intertemporal
budget constraint that the present value of
spending and transfers is bounded by the
present value of future tax collections.
44
Countries
• Countries are collections of individual firms,
households and governments that trade goods
and services as well as financial assets;
• Labor is immobile between countries but mobile
within countries;
• Financial capital is mobile within and between
countries;
• Physical capital is sector and country specific at
any point in time and subject to adjustment
costs over time.
45
Role of Money
• Money is required for transactions between all
agents. There is a technology that combines
money with produced goods and services and
the combined product is what is available in the
market.
• The supply of money is determined by a
central bank in each economy in conjunction
with assumptions about the monetary regime
(e.g. the exchange rate regime).
46
Financial Markets
• Financial markets exist for
– Money
– Government Bonds
– Equity
– Foreign Assets
– Foreign Exchange
• Each financial asset represents a claim over real
resources
– Money over purchasing power
– Bonds are claims over future tax collections
– Equity is a claim over the future dividend streams
– Foreign assets are claims over the future exports
47
Goods and Services Markets
• Households, Firms and Governments trade goods
and services and price for each is assumed to clear
the markets at an annual frequency
48
Factor Markets
– Labor Markets
• Nominal wages are set by different institutional
structures in each country;
• Given the nominal wage and the market prices
for goods and services firms higher labor until
the real wage in each sector equals the
marginal product of labor;
• Aggregate unemployment can result although
over time it is assumed that unemployment
tends to force the nominal wage towards the
labor market clearing level.
49
Factor Markets
– Capital
• once installed physical capital is costly
to move;
• Capital produces a flow of services for
firms that have installed a capital stock
through investment decisions in the
past;
• Investment is subject to rising marginal
costs of installation and depreciation
over time.
50
Factor Markets
– Energy and Materials in GCUBED
• Firms purchase the output of other sectors as
inputs in production;
• Total demand for the materials and energy
sectors is final demand plus demand for
intermediate inputs in each sector;
• In contrast to standard CGE models (which
assume Leontieff fixed coefficients between
intermediate inputs and value added ) there is a
CES production technology which allows
substitution of capital, labor energy and
materials in production.
51
Generating a Baseline Projection
• Usually 2 approaches to policy analysis in the new
generation of global models (CGE or macro)
– Assume at a steady state and analyze deviations
from steady state
– Assume the observed data in a given year is on the
stable manifold of a system dynamically adjusting to a
long run steady state
52
Generating a Baseline Projection
• Given values for all exogenous variables the model is
solved for an equilibrium over time in which all equations
hold given current and expected future variables.
• Underlying the projection is a convergence model for
sectoral productivity growth and exogenous population
projections
• We adjust the model so that we exactly generates the
base year data set (2002).
• Adjustments are made to constants in behavioral
equations
– In arbitrage equations this is equivalent to calculating
risk premia
53
Generating a Baseline Projection
• This generates a baseline projection or forecast from
2002
• We then step the model forward to 2003 adjusting the
information set for 2003 to generate a baseline from
2003 to 2100
54
Running Scenarios
• Once the baseline is generated we run scenarios by
changing exogenous variables or initial conditions
• The information set for markets is critical
– Suppose we expect a shock what does the
anticipation do before the actual shock occurs
• Information sets can be changed by the user
unexpectedly over time so we can ask questions like:
– Suppose we expected a shock but it doesn’t happen
– Suppose we don’t expect a shock and it suddenly
occurs
55
Research Projects/ Challenges
• Model Development
• Estimation at a more detailed level (with Peter
Wilcoxen) to enable more flexible aggregation.
• Incorporating uncertainty in long term projections
(Peter Wilcoxen and various students at Syracuse)
• Incorporating learning by agents (Kang Yong Tan)
• Incorporating demographic variables following
Blanchard/ Yaari/ Weil (Jeremy Nguyen)
• Projecting productivity and convergence (Alison
Stegman)
• Emissions projections for climate predictions
(Alison Stegman) - testing convergence
56
Research Projects/ Challenges
• Model Development
• Modeling developing countries
– Modeling transition economies such as
Vietnam (Hong Giang-Le)
– Tajikstan (Zavkiev Zavkijon)
• Incorporating Infectious diseases (WHO project)
(with Alexandra Sidorenko)
57
Research Projects/ Challenges
• Improving Macroeconomic Dynamics
• Need to integrate the global structural models with the
more data intensive VAR approach
– Using the G-Cubed model to generate restrictions on
multi-country VARS following Pagan and others
• With Dungey, Fry, Pagan
58
Research Projects/ Challenges
• Software development
• Goal is to make the model platform independent so
that users can run in the model in most packages
• Multiple platforms to be supported
– Shifting core software from GAUSS to Ox
– Gempack
– Troll
59
Research Projects/ Challenges
• Training
– Online teaching material for self teaching or course
development
• Beginner, intermediate, advanced
• Applications for non-models
– www.economicscenarios.com
60
Use of Scenarios
• The most effective way to undertake scenario analysis is
with an internally consistent and empirically relevant
framework
• The models form the analytical and empirical basis for
designing alternative scenarios
61
Use of Scenarios
•
•
•
•
•
Ask the question
– What are the likely consequences of the Iraq War?
Design the scenarios that give different insights to the question
– Examine history (Gulf War I, Afghanistan, Vietnam, Korea)
• Wars always cost more than expected
• Costs are more than the fiscal outlays
– Shocks to
• Government spending for the war (US, Aust, UK)
• Government spending for the peace (Europe/Japan)
• Increased global risk
Impose the shocks in a consistent framework (a model)
Interpret the results
Assess the key sensitivities that drive the results
• Do people expect it to be temporary or more permanent?
62
Scenario Examples
www.economicscenarios.com
•
•
•
•
•
•
•
The impact of Sept 11 Terrorist Attacks
The Consequences of WorldCom and Enron Collapses
The Global Economic costs of the Iraq War
The Global Impact of SARS
The Global consequences of Exploding US Fiscal Deficits
The impacts of a Chinese exchange rate revaluation
The Impacts of a Financial Crisis in China resulting from
macroeconomic policy tightening
63
Conclusions
• Many exciting developments in multi-country modeling
due to software and hardware advances
• To some extent the wide divergence in historical
approaches originating from micro versus macro
theoretical schools have converged
• But the access to new datasets and computing
technology have also allowed new approaches to
emerge
• Many important issues with possibly large global
consequences can now be addressed in models - from
the global demographic transition to pandemics
64
Conclusions
• But in the end the existing suite of global models are just
models with many limitations
• Need more work on the construction of consistent
datasets – many existing problems
• Need more work on validation/estimation
• Despite what has been achieved in recent decades, any
model will only be as good as the person using it.
• Most graduate schools do not teach economy-wide
modeling because it is difficult to build an academic
career developing large scale models
65
Background Papers
www.gcubed.com
www.notwrong.com
www.economicscenarios.com
66